Positive semidefinite analytic functions on real analytic surfaces (Q2665608)

This paper is concerned with a solution of Hilbert's 17th Problem for meromorphic functions defined on global real analytic surfaces embedded in \( \mathbb{R}^n\), that is global analytic sets \(X \subset \mathbb{R}^n\) of dimension 2. The author proves that any positive semidefinite function on a global real analytic surface \( X \subset \mathbb{R}^n \) is the sum of 10 squares of meromorphic functions on \( X \). This result is used to establish the Real Nullstellensatz for global real analytic subsets of dimension \( \leq 2 \). The paper begins with a thorough summary of the existing results for Hilbert's 17th Problem in both the global and local real analytic and meromorphic cases for analytic sets of all dimensions before stating the main theorem. Specifically the result proved in this paper is that if \( X \subset \mathbb{R}^n \) is a global real analytic surface and \( f:X \rightarrow \mathbb{R} \) is a positive semidefinite analytic function then there exist analytic functions \( g, f_1, \dots, f_{10} \) defined on \(X\) such that \( g^2 f = \sum_{i = 1}^{10} f_i^2 \) and \( \{g = 0 \} \subset \{f = 0\} \cup B(X) \), where \(B(X)\) is the set of points \( x \in X\) where the local ring \( \mathcal{O}_{X, x} \) is not a normal ring. In addition, it is also shown that in the case when \( X \) is coherent, just 5 squares are sufficient. The result is also valid for global real analytic spaces of dimension 2, and the author states that the proof presented in this paper works verbatim for that case. The proof of the main theorem makes use of the normalization of the complexification of a global real analytic space. The proof proceeds via a careful argument through several lemmas and theorems. In the last step the general case and the coherent case are considered separately, as several simplifications can be made for the latter leading to the tighter bound of 5 on the sum of squares needed. Among the auxiliary results used in the proof, there is one original result that is particularly interesting concerning the approximation of continuous functions on the real part space of a Stein space by invariant holomorphic functions defined on the whole space.